What scite is
scite is a citation analysis service that reads the sentences in which a paper is cited and classifies each one as supporting, mentioning, or contrasting the cited work. Instead of “842 citations”, you see a breakdown — supporting, contrasting, and the large neutral remainder — and you can read the citing sentence itself.
The insight behind it is that a raw citation count answers a question nobody has. Papers are cited to be agreed with, to be used as a method, to be disagreed with, and to be listed in an introduction by someone who did not read them. Collapsing all of that into one number destroys the information a researcher actually wants, which is: did this hold up? scite is the only tool at this scale attempting to restore that distinction, and it is the reason to use it despite everything in the limitations section below.
Why researchers use it
- Contrasting citations, on demand — filter to the papers that disagree and read why, which is the fastest route into a replication controversy.
- The citing sentence, in context — you see the actual claim being made about the paper, not just an entry in a reference list.
- Reference Check for manuscripts — upload a draft and scite scans your bibliography for retracted papers, editorial expressions of concern, and heavily contrasted findings before a reviewer does it for you.
- Retraction awareness — badges flag work whose status has changed since it was cited, which is the failure mode that quietly damages a literature review.
- Browser badges — citation breakdowns appear on journal pages while you are reading, rather than requiring a separate lookup.
Where it fits in a research workflow
scite answers a narrow question at two specific moments. The first is while reading: you have found a striking result and want to know whether the field accepted it. The second is before submission: your bibliography is finished and you want to know whether anything in it has been retracted or seriously challenged.
That second use is the underrated one, and it sits naturally at the end of the writing pipeline — after your citations are assembled in Zotero and before the manuscript leaves your hands. It is complementary rather than competitive with the discovery tools: Connected Papers and Litmaps tell you what exists, Semantic Scholar lets you search it, Consensus summarises what papers conclude, and scite alone tells you what happened to a finding after publication.
Getting started
Twenty minutes, and check access before paying.
- Check institutional access first. A large number of universities subscribe, and personal subscriptions bought unnecessarily are the most common waste here.
- Look up a famous result in your own field and filter to contrasting citations. Reading those statements is the fastest way to calibrate how much the classification is worth.
- Install the browser extension so citation breakdowns appear while you read rather than requiring a detour.
- The step people skip, and the highest-value one: run Reference Check on a finished manuscript before submission. Finding a retracted citation yourself costs an afternoon; having a reviewer find it costs considerably more.
scite vs the alternatives
| Alternative | Does it better | Pick it if |
|---|---|---|
| Semantic Scholar | Free search, citation contexts, and an open API | You want breadth and automation without a subscription |
| Consensus | Summarising what papers conclude about a question | You are orienting on a topic, not auditing a finding |
| Connected Papers | Showing a field’s structure | You need discovery, not verification |
| Elicit | Extracting comparable methods and outcomes into a table | You are building an evidence table |
scite is the only one of these that reads citation intent. If your question is “did this replicate”, nothing else here answers it; for anything else, one of the others is cheaper or free.
Cost, licensing, and your data
Subscription-priced for individuals with academic and student rates, and widely licensed by institutions — check your library before spending, because access through a university is common enough that paying personally is often unnecessary. There is limited free browsing to evaluate the tool.
Coverage depends on publisher agreements, which is the structural constraint to understand: scite can only classify citation statements from full texts it is permitted to process. Fields with less open full text are covered less completely, and absence of contrasting citations may reflect absence of access rather than absence of disagreement. Uploading a manuscript for Reference Check means sending an unpublished draft to a third party, which is worth a moment’s thought if the work is under embargo or commercially sensitive.
The honest review
Strengths. scite produces data that exists nowhere else at this scale. The ability to ask “who disagreed with this, and what did they say” and get an answer in seconds changes how carefully you can read a literature you are new to. For anyone working near a replication-troubled area — much of psychology, parts of biomedicine — it is the single most useful check available.
Limitations. The classification is machine-made and it is wrong often enough to matter. “Contrasting” routinely captures methodological qualification, boundary conditions, and partial replication rather than genuine refutation, and the label carries a rhetorical weight the underlying analysis does not support. Absence of contrasting citations is weak evidence of anything, since it may simply mean the citing full texts were not accessible. Coverage varies sharply by field and publisher. And the price is high for a tool you will consult occasionally rather than daily, which makes personal subscriptions hard to justify.
Verdict. Adopt it if your institution subscribes, and use it habitually while reading and mandatorily before submission. Skip a personal subscription unless you are actively working on replication or evidence appraisal. The condition that flips the answer is institutional access — this is a good tool at zero marginal cost and a questionable one at list price.
Common questions
Is scite free?
No. It is subscription-priced for individuals with academic and student rates, though limited free browsing exists. Many universities hold institutional subscriptions — check with your library before paying, as personal subscriptions are frequently unnecessary.
What does a “contrasting” citation actually mean?
That a machine classifier judged the citing sentence to disagree with the cited work. In practice this ranges from outright refutation to a qualification about a different population or method. Read the statements themselves before drawing any conclusion from the label.
Can scite check my manuscript for retracted references?
Yes — Reference Check scans an uploaded manuscript’s bibliography for retractions, expressions of concern, and heavily contrasted findings. It is the single most valuable thing the service does, and it belongs in your pre-submission checklist.